Download China: Does Government Health and Education Spending

WP/10/16
China: Does Government Health and
Education Spending Boost Consumption?
Steven Barnett and Ray Brooks
© 2010 International Monetary Fund
WP/10/16
IMF Working Paper
Asia and Pacific Department
China: Does Government Health and Education Spending Boost Consumption?
Prepared by Steve Barnett and Ray Brooks1
Authorized for distribution by Nigel Chalk
January 2010
Abstract
This Working Paper should not be reported as representing the views of the IMF.
The views expressed in this Working Paper are those of the author(s) and do not necessarily represent
those of the IMF or IMF policy. Working Papers describe research in progress by the author(s) and are
published to elicit comments and to further debate.
Consumption in China is unusually low and has continued to decline as a share of GDP over
the past decade. A key policy question is how to reverse this trend, and rebalance growth
away from reliance on exports and investment and toward consumption. This paper
investigates whether the sizable increase in government social spending in recent years
lowered precautionary saving and increased consumption. The main findings are that
spending on health, but not education, had an impact on household behavior. The impact,
moreover, is large. A one yuan increase in government health spending is associated with a
two yuan increase in urban household consumption.
JEL Classification Numbers: H31, H51, H52
Keywords: China, saving, consumption, health, education
Author’s E-Mail Addresses: [email protected]; [email protected]
1
We would like to thank Jianxiong He, Vivek Arora, Tarhan Feyzioglu, Nigel Chalk, Papa N’Diaye,
Nathan Porter, Francis Vitek, and seminar participants at the People’s Bank of China for helpful comments and
suggestions, as well as Kessia De Leo and Imel Yu for their assistance in preparing this paper. All remaining errors
are ours.
2
Contents
Page
I. Introduction ............................................................................................................................3 II. Consumption and Saving in China: Stylized Facts ...............................................................4 A. Consumption is Falling .............................................................................................4 B. More Stylized Facts ...................................................................................................7 III. Reducing Precautionary Saving: A Role For Public Spending? ..........................................8 A. Urban Households .....................................................................................................8 B. Rural Households ....................................................................................................10 C. Robustness Checks ..................................................................................................10 IV. Conclusion .........................................................................................................................11 Figures
1. Consumption in China: Low and Falling ...............................................................................4 2. Urban and Rural Saving Rates ...............................................................................................5 3. Urban and Rural Income ........................................................................................................5 4. Indicators of Urban and Rural Consumption and Income .....................................................6 5. Household Income and GDP per Capita ................................................................................7 6. Urban Household Saving Rate by Income Group..................................................................7 7. Health and Education Spending .............................................................................................9 Tables
1. Urban Households: Saving and Government Spending.......................................................12 2. Rural Households: Saving and Government Spending ........................................................12 References ................................................................................................................................13
3
I. INTRODUCTION
Consumption in China is unusually low and has continued to decline as a share of GDP over
the past decade. A key policy question is how to reverse this trend, and rebalance growth
away from reliance on exports and investment and toward consumption. The collapse in
global demand—and corresponding decline in China’s export growth—is making
rebalancing even more urgent.
China’s low household consumption, or equivalently high saving, is often linked to
precautionary motives. Government health, education, and pensions systems are underdeveloped, leaving individuals to bear a large share of the costs. As a result, households build
up saving to cover these expenses, as well as to self-insure against uncertainty, especially
regarding future health and pension needs. Of course, to explain both the low and declining
consumption rate, precautionary motives must be large and also rising over time. Indeed,
Chamon and Prasad (2008) argue this is the case, as the breaking of the “iron rice bowl” with
the reforms in the 1990s—especially to state-owned enterprises—led to a breakdown of
existing systems and increased household uncertainty.
An obvious channel for boosting consumption, therefore, is to reduce precautionary saving.
Indeed, this has in many respects become a consistent view in the literature. To cite some
examples, IMF (2007) and Blanchard and Giavazzi (2005) both emphasize reducing
precautionary saving as one of the key components for rebalancing growth. The goal of this
paper is to explore the empirical relationship between government health and education
spending and household saving.
The main conclusion that comes out of the data is that government spending on health, but
not on education, has an impact in reducing urban household saving. The impact, moreover,
is large. A 1 yuan increase in government health spending was associated with a 2 yuan
increase in household consumption. Total (household plus government) consumption could
thus increase by as much as 3 yuan depending on the extent that government health spending
takes the form of consumption instead of transfers. For rural households, on the other hand,
the evidence is more mixed. Increased government health spending in rural areas appears
only to have an impact on savings in the higher-income provinces.
We focus narrowly on the issue of the role of government spending, and do not attempt to
answer the broader question of what is driving consumption behavior. Household income, as
highlighted in Aziz and Cui (2007), is an even more important driver of consumption’s
falling GDP share than the saving rate. This would suggest that policies to boost household
income would also bear fruit, particularly if they are combined with policies, such as
improving the healthcare system, that are effective in reducing the household saving rate.
The remainder of this paper proceeds as follows. In the next section, we draw some stylized
facts about consumption and saving in China. In section III, we use provincial household data
4
to examine econometrically the relationship between saving and government health and
education spending. The last section reviews the main findings.
II. CONSUMPTION AND SAVING IN CHINA: STYLIZED FACTS
A. Consumption is Falling
Consumption as a share of GDP in China has steadily declined, and is quite low relative to
other countries. This is clear in Figure 1, where the top panel shows total consumption as a
share of GDP for a cross-section of countries. Since there is some substitutability between
public and private consumption—government provision of free education, for example,
means less need for household education spending—it is useful to look at total consumption.
The second panel shows how the share of household consumption has been falling over time.
Mechanically, this has been offset by rising investment and more recently by an expansion in
net exports.
Figure 1. Consumption in China: Low and Falling
100
Selected Countires: Consumption Expenditure
(In percent of GDP, avereage for 2003-07)
100
90
90
80
80
70
70
India
60
Japan
Thailand
Korea
60
Malaysia
Singapore
Ireland
50
50
China
40
0
10,000
60
20,000 30,000 40,000 50,000
GDP per capita (US$, 2007)
60,000
40
70,000
China: Consumption and Investment, 1990-2008
(In percent of GDP)
50
60
50
40
40
30
30
20
20
10
10
0
0
-10
Household consumption
Government cons.
Net exports
Gross fixed capital
-10
-20
-20
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
5
Saving rates have risen, helping explain the
decline in the consumption ratio (Figure 2).
From the household survey data it is clear
that rural and urban households have
followed quite a different path. The urban
saving rate has steadily risen over time
whereas rural saving has been more
volatile, but has also risen. Urban
households, however, account for the lion’s
share of economic activity and so are
instrumental in the overall increase in
aggregate saving (see below).
In addition to a higher saving rate,
household income has also been a major
factor helping to explain the decline in the
consumption ratio. In per capita terms, both
urban and rural household incomes have
fallen steeply over time, failing to keep pace
with GDP growth (Figure 3). Households
are getting a declining share of the value
added pie (see Aziz and Cui, 2007). From
the early 1990s to 2008, urban household
per capita income dropped from nearly
90 percent of GDP per capita to just under
70 percent.
35
Figure 2. China: Urban and Rural Saving Rates, 1991-2008
(In percent of income)
35
30
30
25
25
20
20
15
15
Rural
Urban
10
10
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
Sources: CEIC; and staff calculations.
40
95
Figure 3. China: Urban and Rural Income, 1991-2008
(In percent of GDP per capita)
90
35
85
30
80
25
75
Rural, net (LHS)
20
70
Urban, disposable (RHS)
15
65
1992
1994
1996
1998
2000
2002
2004
2006
2008
Sources: CEIC; and staff calculations.
Note: Left and right axis each span 25 percentage points, so slopes are comparable.
This decline in household’s share of income has been the main culprit behind the drop in
consumption (Figure 4). For urban households, roughly 60 percent of the decline in
consumption since the early 1990s is attributable to a drop in disposable income as a share
GDP. For rural households, falling income plays a larger role and explains around
three-fourths of the decline since the early 1990s.
6
Figure 4. Indicators of Urban and Rural Consumption and Income
Urban consumption as a share of income is falling whether
measured by national accounts or household survey data.
100
China: Urban Consumption Ratio,
1991-2008 1/
90
The same holds for rural consumption data, though
rural consumption is a smaller share of income.
100
90
80
80
70
70
60
60
National accounts
Household survey
50
40
1994
1996
1998
China: Rural Consumption, 1991-2008
(Nominal per capita, in percent of GDP per capita)
35
2000
2002
2004
2006
2008
25
25
20
20
15
15
10
National accounts
10
5
Household survey
5
0
0
1992
1994
1996
1998
2000
2002
2004
2006
2008
Sources: CEIC; and staff calculations.
There is no clear sign of consumption smoothing in
the urban household survey data…
14
35
30
1/ Nominal per capita urban consumption in percent of nominal per capita GDP.
Sources: CEIC; and staff calculations.
16
40
30
50
40
1992
40
China: Urban Consumption and Income Growth, 1991-2008
(Per capita real growth, in percent)
…nor for rural households.
16
14
18
16
China: Rural Consumption and Income Growth, 1991-2008
(Per capita real growth, in percent)
14
18
16
14
12
12
10
10
10
8
8
8
8
6
6
4
4
2
2
0
6
6
4
Consumption (Survey)
4
2
Net income
2
0
0
-2
Consumption (SNA)
0
1992
1994
1996
1998
2000
2002
2004
2006
Urban income (consumption) per capita is about 3 times
bigger than rural income (consumption) per capita…
China: Urban-Rural Gap, 1991-2008 1/
3.8
100
3.6
90
2.6
30
2.4
Consumption
2.4
20
2.2
10
2.0
2006
1/ Ratio of urban to rural in per capita terms, based on household survey data.
Sources: CEIC and staff calculations.
2008
(In percent of total)
100
90
40
Income
2004
China: Urban Population and Consumption, 1990-2008
50
2.6
2002
2008
60
40
2000
2006
50
2.8
1998
2004
70
2.8
1996
2002
60
3.0
1994
2000
80
3.0
1992
1998
70
3.2
2.0
1996
80
3.2
2.2
1994
…but urbanization has dramatically increased the share
of urban consumption in total consumption.
3.4
3.4
10
Sources: CEIC; and staff calculations.
Note: "SNA" refers to national accounts, and "survey" to urban household survey.
Sources: CEIC; and staff calculations.
Note: "SNA" refers to national accounts, and "survey" to urban household survey.
3.6
12
-2
1992
2008
3.8
Consumption (Survey)
Net income
Consumption (SNA)
12
30
Urban share of consumption 1/
20
Population 2/
10
0
0
1992
1994
1996
1998
2000
2002
2004
1/ Urban consumption as a share of total household consumption.
2/ Urban population as a share of total population.
Sources: CEIC; and staff calculations.
2006
2008
7
B. More Stylized Facts
Some other characteristics help put the trends in saving, consumption, and income into
perspective:
First, real growth in household income and
consumption has been fast (Figure 5).
However, real GDP growth has been even
faster. National accounts data indicate that
real household consumption (CPI deflated)
has grown at a remarkable 8¾ percent
average annual rate over the past 10 years.
GDP growth, however, has been even
faster at 9¾ percent. Household income
has also risen quickly but not as fast as the
pace of GDP growth.
16
16
Figure 5. China: Household Income and GDP per Capita, 19912008
14
14
(Real growth, in percent)
12
12
10
10
8
8
6
6
Rural
4
4
GDP
Urban
2
2
0
0
1991
1993
1995
1997
1999
2001
2003
2005
2007
Sources: CEIC; and staff calculations.
Second, the data lack any obvious signs of consumption smoothing. The volatility of
consumption growth is greater than that of income growth. This is true for both urban and
rural households. For the urban data, both income and consumption jump up in 2002, which
Chamon and Prasad (2008) note was when improvements were made to the household survey
methodology. This suggests the spike in the data that year could merely represent a break in
the methodology.
Third, urban consumption accounts for three-fourths of household consumption, a share that
has risen steadily in line with the pace of urbanization. Indeed, rural consumption accounts
for most of the decline in the consumption to GDP ratio. Less obvious, however, is that the
share of urban population has also risen significantly. Expressed in per capita terms, to
control for urbanization, the urban-rural gap has widened over the sample period, but has
stabilized more recently. The gap, however, remains large as urban households earn and
spend roughly three times as much as rural ones. This is also relevant to the extent that
saving behavior would be expected to be dependent on income levels.
Fourth, as would be expected saving is
positively related to income. The trends in
saving, however, also differ. Lower
income groups have seen relatively little
increase in saving rates and indeed saw a
decline in their savings in the early 2000s
(Figure 6). In contrast, higher income
groups have had gradually increasing
saving rates with the pace of that increase
starting to accelerate around 2003.
45
35
Figure 6. China: Urban Household Saving Rate by Income Group 1/
(In percent)
25
15
5
-5
-15
1997
1999
2001
2003
2005
2007
1/ Moving upward lines are: Lowest: poor; Lowest; Low; Lower middle; Middle; Upper middle;
High; and Highest.
Sources: CEIC and staff calculations
8
III. REDUCING PRECAUTIONARY SAVING: A ROLE FOR PUBLIC SPENDING?
The econometric section examines whether increases in government spending on health and
education have an impact on savings. Provincial data are pooled to exploit variations in
provincial spending on health and education, and differences in saving rates. Data from
1994–2007 are used, but a smaller sample of 2003–07 is also employed to check robustness
and control for the possibility of structural changes.
The analysis looks at urban and rural households separately. As noted above, the trends in
their saving behavior differs as does their income levels, making it likely that their response
to increases in government spending would not be the same. Ideally, therefore, the
government spending variables used—for education and health spending—would breakdown
spending into that undertaken in urban and rural areas. The provincial government spending
data, however, do not distinguish between these types of spending. So the differences in
results found between urban and rural behavior should be treated with some caution since it
could merely reflect differences in how government spending has been allocated.
A. Urban Households
The results suggest a statistically significant relationship between health spending and urban
saving (Table 1). The parameter estimate is highly statistically significant and fairly stable
across the two samples and with the inclusion of education spending. The estimate is around
-2, which suggests that each 1 yuan of government health spending results in a 2 yuan
decrease in saving—or, equivalently, a 2 yuan increase in consumption. This is a strong
impact, as it would imply that a 1 percent of GDP increase in government health spending
would boost private consumption by 2 percent of GDP and yield a total demand effect of
3 percent of GDP for every 1 percent of GDP increase in health spending.
The impact of government health spending on saving is larger than might have been
expected. In this first step, government health spending simply substitutes for private health
spending, which would cause the saving rate to increase. That is, for each extra yuan the
government spends on health individuals could spend 1 yuan less, freeing up that 1 yuan to
be allocated to additional saving and consumption—but the net impact, provided at least part
of the freed up income is saved, would be an increase in the saving rate. The negative sign,
therefore, indicates that government spending is not substituting for private spending and
suggests that it is instead reducing the need for saving. That is, higher government health
spending seems to reduce the need for precautionary saving and frees up households ability
to spend on other goods and services.
Private health care spending by urban households has fallen in recent years, suggesting that
higher government spending on health care has indeed freed up resources for households for
other spending or saving. Urban health care spending (from the household survey) fell by
9
more than ½ percent of urban household consumption over the past four years, as
government spending on health care more than doubled to the equivalent of 3 percent of total
household consumption (Figure 7). A similar picture emerges from the national accounts
data, which show higher urban health care spending, but nonetheless a decline in recent
years. Rural healthcare spending, however, increased over the past four to five years by ½–
¾ percent of rural household consumption (based on survey and national accounts data).
Figure 7. China: Health and Education Spending
1,800
4.00
3.50
China: Government Health and Education Expenditure
(As percent of GDP)
1,600
Government Real Health and Education Spending
(RMB, 2007 prices)
1,400
3.00
1,200
2.50
Education
1,000
2.00
Education
800
1.50
600
1.00
400
Health
0.50
0
0.00
2000
2001
2002
2003
2004
2005
2006
2007
2000
2008
2001
2002
2003
2004
2005
2006
2007
2008
12.0
10.0
9.0
Health
200
Heath Care Spending
(As percent of private consumption, from household survey)
8.0
Urban
7.0
10.0
Heath Care spending
(As percent of private consumption, from national accounts)
Urban
8.0
6.0
Rural
6.0
5.0
Rural
4.0
4.0
3.0
2.0
Government
2.0
Government
1.0
0.0
0.0
2000
2002
2004
2006
2008
2004
2005
2006
2007
Sources: CEIC data; and authors’ estimates.
The results for education are, however, quite different. There is little evidence that higher
government education spending decreases saving. The coefficient is never statistically
significant, and the sign actually switches between the full sample and the latter period. The
lack of a statistically significant relationship could reflect that much of the increase in
government education has been toward primary and secondary education, whereas the saving
is more targeted toward higher education. Moreover, the increase in government spending
may not have kept pace with the rise in expected household education spending. Education
spending per capita is more than five times larger than health spending and has also grown
rapidly over the past four years.
10
B. Rural Households
For rural households, there is little evidence that higher health spending reduces
precautionary saving (Table 2). The coefficient estimate on health is never statistically
significant, and for that matter is always positive, which would suggest that higher
government health spending actually increases saving. This may reflect the fact that many of
the rural households are subsistence consumers and would be expected to save a substantial
part of the resources freed up by the government providing more health care coverage. It
could also be that rural areas are receiving a smaller share of provincial health spending and
so it is having much less effect than in the urban areas or the increase in public spending has
been small relative to expected rural health spending needs.
Likewise, there is little evidence that an increase in education spending has an impact on
saving of rural households. Again, the coefficient is always positive, and, while not
statistically significant at conventional levels, is quite close to being significant in the
2003-07 sample. This would be consistent with increase government spending on education
simply substituting for private spending—and the coefficient of near unity would suggest a
nearly one-for-one impact.
C. Robustness Checks
The coastal provinces have developed faster, have higher income per capita, and tend to be
more urbanized. The regressions are repeated allowing the impact to be different in the eight
highest income provinces.2 This helps check the robustness of the results more generally, but
also examines whether there are systematic differences in the highest income provinces.
The results for rural households suggest that government health spending in high income
provinces, unlike for the whole sample, is associated with a decline in rural household
savings. The magnitude, moreover, is similar to the -2 found in the urban regressions
reported above. This could reflect a variety of factors, including that rural households in high
income provinces are wealthier (and thus resemble urban households elsewhere) or that they
are a larger beneficiary of government health spending.
The results for urban households, however, do not provide strong evidence that the impact of
health spending differs in higher income provinces. The interaction term is not usually
statistically significant, though it is always negative, which would suggest that the impact
could be marginally stronger in higher income provinces.
Finally, the results for the impact of government health spending on urban household saving
hold up to other robustness checks. Different explanatory variables are included to gauge the
2
These are Beijing, Tianjin, Shangdong, Jiangsu, Fujian, Guangdong, Zhejiang, and Shanghai.
11
impact on the statistical significance and magnitude of the parameter estimate on health
spending. Variables that are added, in various combinations, include real income growth,
inflation, and unemployment. The results (not reported) did not change substantially in either
the whole sample or the 2003–07 sample.
IV. CONCLUSION
The main finding of this paper is fairly robust. Higher government health spending reduces
urban household saving and suggests that broadening coverage of public health care could
have an important effect on household precautionary savings. The magnitude of the impact,
moreover, is quite large and suggests that each additional yuan in government health
spending boosts urban consumption by 2 yuan.
For rural households, with the exception of those in the higher income provinces, there is,
however, no evidence of a relationship between government health spending and saving.
There was also no evidence that higher government education spending has an impact on
either urban or rural saving. This is not entirely surprising, as the precautionary saving
motives are likely much higher for health than education. In both cases, an increase in
government spending has competing effects on saving. First, it could increase saving by
substituting public for private provision of the services, thereby freeing up household
resources of which some would be saved. To reduce saving, government education or health
spending would need to have an additional effect, such as reducing precautionary saving
motives. Given that a large part of the savings are being generated by older generations, who
are more likely to need to save for healthcare costs rather than education for their family
members, it is sensible to expect that improvements in public health care are likely to have a
more potent effect on household saving behavior than expanding publicly provided
education.
Although past rural health care spending appears to have had little impact on consumption,
the government’s new health reform strategy for 2009–11 has the potential to improve health
outcomes and raise rural consumption. The government has given priority to the health
sector, along with other social sectors, such as education and social protection, in the
11th five-year plan. A comprehensive study by the World Bank (2009) suggests that while
progress is being made to improve China’s rural health system, including by marching
quickly toward universal coverage for rural areas, a number of challenges remain.
Importantly, the study notes that the inequalities in China’s health system reflected, at least in
part, inequalities in government health spending. They note that government health
expenditure disproportionately benefited the better off. This reflects the fact that over half of
general government health spending supports urban insurance schemes whose members are
disproportionately from higher income groups, even within urban areas. The reorientation of
government spending toward the poorer rural areas and increased efficiency in the delivery
of public health care will be key factors in improving health outcomes and impacting
consumption behavior.
12
Table 1. Urban Households: Saving and Government Spending
(1)
Sample: 1994-2007
(2)
(3)
(4)
(5)
Sample: 2003-07
(6)
(7)
(8)
Health
Estimate
(Standard error)
[P-val]
-2.10
(0.72)
[0.00]
...
...
...
-1.92
(0.86)
[0.03]
...
...
...
-1.94
(0.60)
[0.00]
...
...
...
-2.06
(0.58)
[0.00]
...
...
...
Education
Estimate
(Standard error)
[P-val]
...
...
...
-0.78
(1.10)
[0.48]
-0.44
(1.07)
[0.68]
...
...
...
...
...
...
0.42
(1.20)
[0.73]
0.66
(1.07)
[0.54]
...
...
...
Health & education
Estimate
(Standard error)
[P-val]
...
...
...
...
...
...
...
...
...
-0.90
(0.63)
[0.16]
...
...
...
...
...
...
...
...
...
-0.41
(0.89)
[0.64]
0.24
285
0.22
304
0.25
285
0.24
285
0.19
150
0.18
150
0.19
150
0.18
150
R-squared
# Obs.
Sources: CEIC; and staff estimates.
Note: All variables are in first differences. The dependent variable is the saving rate, and government
spending variables are per capita spending expressed as a share of per capita urban disposable
income (lagged one period). Pooled Provincial data are used, with fixed and time effects.
Table 2. Rural Households: Saving and Government Spending
(1)
Sample: 1996-2007
(2)
(3)
(4)
(5)
Sample: 2003-07
(6)
(7)
(8)
Health
Estimate
(Standard error)
[P-val]
0.51
(0.59)
[0.39]
...
...
...
0.22
(0.58)
[0.70]
...
...
...
0.37
(0.67)
[0.58]
...
...
...
0.06
(0.64)
[0.93]
...
...
...
Education
Estimate
(Standard error)
[P-val]
...
...
...
0.45
(0.36)
[0.22]
0.49
(0.38)
[0.20]
...
...
...
...
...
...
0.91
(0.61)
[0.14]
0.90
(0.54)
[0.10]
...
...
...
Health & education
Estimate
(Standard error)
[P-val]
...
...
...
…
...
...
...
...
...
0.39
(0.29)
[0.18]
...
...
...
...
...
...
...
...
...
0.53
(0.47)
[0.27]
0.36
285
0.37
304
0.36
285
0.36
285
0.31
150
0.32
150
0.32
150
0.32
150
R-squared
# Obs.
Sources: CEIC; and staff estimates.
Note: All variables are in first differences. The dependent variable is the saving rate, and government
spending variables are per capita spending expressed as a share of per capita urban disposable
income (lagged one period). Pooled Provincial data are used, with fixed and time effects.
13
REFERENCES
Aziz, Jahangir, and Li Cui, 2007, “Explaining China’s Low Consumption: The Neglected
Role of Household Income,” IMF Working Paper 07/181 (Washington: International
Monetary Fund).
Blanchard, Olivier, and Francesco Giavazzi, 2005, “Rebalancing Growth in China: A ThreeHanded Approach,” MIT Department of Economics Working Paper No. 05-32
(Cambridge: Massachusetts Institute of Technology).
Chamon, Marcos, and Eswar Prasad, 2008 “Why are Saving Rates of Urban Households in
China Rising?” IMF Working Paper 08/145 (Washington: International Monetary Fund).
International Monetary Fund, “China’s Difficult Rebalancing Act,” IMF Survey Online,
September 12, 2007. Available via Internet:
http://www.imf.org/external/pubs/ft/survey/so/2007/CAR0912A.htm
Wangstaff, Adam, Magnus Lindelow, Shiyong Wang, and Shuo Zhang, “Reforming China's
Rural Health System,” July 2009 (Washington, DC: The World Bank).